xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200385Ka 4500
controlfield tag 001 002069506
007 cr mnu|||uuuuu
008 100422s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003285
Afro-Colombian welfare :
b an application of Amarty Sen's Capability Approach using multiple indicators multiple causes modeling MIMIC
h [electronic resource] /
by Paula Lezama.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 72 pages.
Thesis (M.A.)--University of South Florida, 2009.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
ABSTRACT: This research analyzes welfare conditions of Afro-Colombians vis--vis non Afro-Colombians using Amartya Sen's Capability Approach as the theoretical framework, and the latent variable modeling as the empirical method. Multiple Indicators Multiple Causes (MIMIC) models are estimated using data from the Colombian Quality of Life Survey, 2003. Two latent constructs, namely, 'knowledge' and 'being adequately sheltered', represent the two 'Capability' dimensions to be analyzed. Ethnic background appears to have a consistently negative influence after controlling statistically by a set of relevant variables (e.g. being poor, area, marital status, age and gender, among others) included in the models as exogenous "causes" or "determinants" of each capability dimension.This implies that the capability set or the freedom an Afro-descendant enjoys in achieving the life he or she wants in terms of 'knowledge' and 'shelter' is consistently lower than that of a non Afro-descendant (whites and mestizos). As a consequence, achieved welfare or functioning achievement as expressed in terms of aspects such as years of education or dwelling conditions in the household is and would be consistently lower for individual members of this ethnic group. This evidence points toward the proposition that embedded patterns of racial discrimination are limiting Afro-Colombian capabilities and individual agency, beyond income levels or even access to educational resources. Hence, from a capability perspective removing racial discrimination must be an explicit objective of developmental policy.Accordingly, national policy must be directed not only to improving access for Afro-Colombians to resources and economic wellbeing, as traditional analysis of class disparity argues, but also toward the nurturing and expansion of the real freedom they have to pursue the goals they value. Thus, development policy in Colombia must altogether work toward the improvement of resource access for Afro-descendants and toward the creation of specific mechanisms to enforce the judicial instruments to fight against racial discrimination. These laws and judicial mechanisms were created to open spaces for political, social and economic participation for this population group on an equal basis, as their fellow citizens of non African descent, and are yet to be fulfilled.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Jorge Nef, Ph.D.
Latent class modeling
x Latin American, Caribbean and Latino Studies
t USF Electronic Theses and Dissertations.
Afro-Colombian Welfare: An Application of Amartya Sens Capability Approach Using Multiple Indicators Multiple Causes Modeling MIMIC by Paula A. Lezama A thesis submitted in partial fulfillment of the requirements for the degree of Masters of Arts Institute for the Study of Latin American and the Caribbean College of Arts and Sciences University of South Florida Major Professor: Jorge Nef, Ph.D. Professors: Bernd Reiter, Ph.D. Enjung Choi, Ph.D. Date of Approval: November 03, 2009 Keywords: Race, Racial Discrimination, Capabilities, Functionings, Latent Class Modeling Copyright 2009, Paula Lezama
Dedication This thesis is dedicated to my husband and daughter, and to my family in Colombia. To them I am in eternal debt for their love a nd unconditional support. They have been by my side during the good and not so good times, and I am very grateful for being part of their lives as they are part of mine. It is also dedicated to all my friends because they have been a permanent source of joy, friendship and learning. Finally, it is dedicated to my profe ssors at the Universidad del Valle, Cali-Colombia and at th e University of South Florid a for sharing their knowledge and passion with me.
Acknowledgments It is a pleasure to thank those who have made this thesis possible. I want to thank Jorge Nef my major professor who has made hi s unconditional support av ailable in many different ways. I also want to thank Be rnd Reiter and Enjung Choi for guiding and encouraging me in this long journey, and to the Institute for the St udy of Latin America and the Caribbean at the University of Sout h Florida for their con tinuous practical and financial support. In addition, I want to express my gratitude to the International Human Development and Capability Association for the opportunity of being part of their summer school 2009, because it was a definitive experience in the process of finishing this Masters Thesis. I am very grateful for the comments on the or iginal proposal made by Sergio Olivieri, to Nini Johana Serna for he r aid on the data base processing stage of this paper, and to Hillary Wade for her help on the editing stage. Finally, my thanks go to Heidi Taylor for her continuous encouragement and friendship since I became a student at the University of South Florida.
Table of Contents List of Tables ....................................................................................................................... i List of Figures ............................................................................................................... ...... ii Abstract ...................................................................................................................... ........ iv Chapter 1. Introduction ....................................................................................................... 1 Chapter 2. Background: Afro-Col ombian Living Conditions ............................................ 6 Chapter 3. Theory And Model Building ........................................................................... 18 Race, Racial Discrimination And Poverty ......................................................... 18 Amartya Sens Capability Approach .................................................................. 22 Chapter 4. Operationalizing Amartya Sen s Capability Approach: A Mimic Model Applied To Colombia ............................................................................................. 31 Data And Conceptual Clarifications .................................................................. 31 Mimic Modeling ................................................................................................. 34 Path Diagram For Mimic Model Of Latent Capabilities .................................... 48 Chapter 5. Results ............................................................................................................ 50 Chapter 6. Conclusions ..................................................................................................... 62 References .................................................................................................................... ..... 65 Appendix A .................................................................................................................... ... 72 i
ii List of Tables Table 1 Distribution of the Colombian Population by Area and Ethno-Racial Characterization ................................................................................................. 10Table 2 Percentages of Poor and Indige nt Populations for 2003, by Urban-Rural Areas* ................................................................................................................ 12Table 3 Percentages of Poor Populations by Urban/Rural and Ethnic Membership base on the QLS 2003 ........................................................................................ 12Table 4 Educational Level by Ethnic Characteri zation based on the 2003 Quality of Life Survey .................................................................................................... 14Table 5 Variable Desc ription and Measurement ............................................................ 41Table 6 MIMIC model of Knowledge Capability Achievement: Factor Loadings of the Structural Model ...................................................................................... 50Table 7 MIMIC Model of Knowledge Capability Achievement. Factor Lo adings for the Measurement Model ............................................................................... 52Table 8 MIMIC model of Housing Capability Achievem ent Factor Loadings of the Structural Model ...................................................................................... 56Table 9 MIMIC Model for Shelter Capability Achievement. Factor Lo adings for the Measurement Model ............................................................................... 58
List of Figures Figure 1 Path Diagram for a general MIMIC model ...................................................... 49Figure 2 Path Diagram for the Knowledge Capability ................................................... 55Figure 3 Path Diagram for the Shelter Capability .......................................................... 59Figure 4 Distribution of Afro-Colombian Population by their Location on the National Territory ............................................................................................. 72 iii
Afro-Colombian Welfare: An Application of Amartya Sens Capability Approach Using Multiple Indicators Multiple Causes Modeling MIMIC Paula A. Lezama ABSTRACT This research analyzes welfare conditions of Afro-Colombians vis--vis non AfroColombians using Amartya Sens Capability Approach as the theoretical framework, and the latent variable modeling as the empirica l method. Multiple Indicators Multiple Causes (MIMIC) models are estimated using data fr om the Colombian Quality of Life Survey, 2003. Two latent constructs, namely, knowle dge and being adequately sheltered, represent the two Cap ability dimensions to be analyzed. Ethnic background appears to have a consistently negative in fluence after controlling statisti cally by a set of relevant variables (e.g. being poo r, area, marital status, age and gender, among others) included in the models as exogenous causes or determinants of each capability dimension. This implies that the capability set or the freedom an Afro-descendant enjoys in achieving the life he or she wants in terms of knowledge an d shelter is consiste ntly lower than that of a non Afro-descendant (whites and mestizos). As a consequence, achieved welfare or functioning achievement as expressed in terms of aspects such as years of education or iv
v dwelling conditions in the household is and w ould be consistently lower for individual members of this ethnic group. This evidence points toward the proposition that embedded patterns of racial discrimination are limiting Afro-Colombian capabilities and individual agency, beyond income levels or even access to educational resources. Hence, from a capability perspective removing racial discrimination must be an explicit objective of developmental policy. Accordingly, national policy must be directed not only to improving access for Afro-Colombians to resources and economic wellbeing, as traditional analysis of class disparity argues, but also towa rd the nurturing and expansi on of the real freedom they have to pursue the goals they value. Thus, development policy in Colombia must altogether work toward the improvement of resource access for Afro-descendants and toward the creation of specific mechanisms to enforce the judicial instruments to fight against racial discrimination. These laws and judicial mechanisms were created to open spaces for political, social and economic participation for this population group on an equal basis, as their fellow citizens of non Af rican descent, and are yet to be fulfilled.
Chapter 1. Introduction Amartya Sens Capability Approach has emer ged as an alternative framework to longestablished economical approaches that seek to analyze individua l welfare, poverty and human development (Kuklys 2005; Clark 2005). He develops the Capability Approach, which identifies two elements, functionings and capabilities to analyze individual welfare. Functionings represent the achieved welfare of an indivi dual: what a person manages to be or do in the different di mensions of human life. These beings and doings vary in a wide range of aspects from being adequately nourished and in good health to more complex achievements like having self-confidence. Capabilities then, are the set of choices people have or the freedom to choose the life they want. They express the set of available options to the exerci se of individual freedom (Sen 1985, 1999). For Sen, freedom is the core aspect of human de velopment, and is determined by evaluating the extent to which a person is able to live the life he or she values and has reason to value; in this manner we can really as sess a persons we llbeing. The author conceptualizes human development as the process of expanding the real freedoms that people enjoy. (Sen 1999, p 3) and institu tional and social arrangement must procure this expansion. Following Sens Capability Approach this research analyzes the wellbeing of Afro-Colombians in re lation to non Afro-Colombians in order to understand how ethnic 1
and racial characteristics aff ect the achievement of welfare for individual members of this ethnic minority. The analysis is made by implementing a Multiple Indicators Multiple Causes (MIMIC) Model to assess capabilitie s in two dimensions of human welfare, knowledge and being adequately sheltere d, as latent constructs (Joreskog and Goldberger 1975). The aim of MIMIC is to capture how a set of exogenous variables (e.g. ethnicity, gender, age, or being poor) influence a given capability dimension (e.g. knowledge), which is assessed as an unobserve d variable. In turn, that given capability determines specific outcomes or individua l functioning achievements (e.g. achieved years of education). These individual functioning ach ievements are the set of indicators of the given latent construct. A latent variable re presents a complex conceptual construct that cannot be measured directly and has to be inferred or approached through a set of observed variables called indicators. Given the multidimensional character of Capabilities as comprising the set of choices an individual has, it is very diffi cult to measure them directly; thus, we have to rely on a set of observed Functionings or achieved welfare variables to infer the behavior of individual capabilities. For instance, we do not have certainty about the choices a person have when deciding which level of education wants to achieve, all we have is the actual level of education achieved such as a college degree. Thus, by analyzing the interplay of variables such gender, age, ethnicity and income, etc. and the achieved levels of education, one can make some inferences about the individual capability that person enjoys when deciding that a college degree is what he or she wanted. Since its inception the Capability Approach has been considered abstract reasoning for its difficulty in being im plemented (Comim 2001). These ideas are 2
philosophical and ethical and there are few empirical applications to measure them. New methods are being increasingly used to try to measure such intrinsic ideas and the Multiple Indicators Multiple Causes (MIMIC) m odel is one such quantitative, statistical technique. MIMIC models were first introduced by Joreskog and Golderberg (1975), and have constituted a common methodology in disciplines such as psychology, econometrics and marketing among others. Recently, author s such as Kuklys (2005), Di Tommaso (2007), Krishnakumar and Ballon (2008), have s hown that latent variable modeling, of which the MIMIC model is part, is an appealing framework when assessing Capabilities and Functionings with existing statistical da ta. The study uses the 2003 Quality of Life Survey (ECV 2003 for its acronym in Spanish)1 implemented by the National Department of Statistics in Colombia, a nd calculations where performed using the Statistical Analysis System (SAS), and its procedure Covariance Analysis of Linear Structural Equation (CALIS). Historically, Afro-Colombians have been marginalized from the benefits of modernity and democracy. Socially excluded, they have the worst living conditions in comparative terms to the rest of the populat ion (Arocha 1998; Urre a et al. 2005; Escobar 2003). They have the highest illiteracy rate s in the country, 15.2% against 8.1% of non Afro-Colombians, and health and housing indicators follow the same trend (Urrea et al. 2005). Under standard welfare analysis of grou p disparities, the deprived conditions of Afro-Colombians have been c onsidered the accumulated resu lt of low income and class disparities. However, the persistence and ma gnitude of the racial gaps in the country suggest that practices of racial discrimina tion are limiting Afro-Colombians individual 1. For practical purposes from now on the survey woul d be refer as ECV 2003 as its acronym in Spanish. 3
agency. Consequently, it can be argued that systematical lower capability and functioning achievement for Afro-Colombians in relation to non Afro-Colombians is at large extent due to unyielding social pr actices of racial discrimi nation and not only to class differences as has been traditionally ar gued by social stratif ication analysis. As a consequence, this thesis argues th at social stratifica tion theory and its concentration on class disparities as the main source of inequality, and income as the main unit of analysis, is limited when assessing r acial inequalities. It is asserted that this shortcoming can be overcome by the inclusion of racial discrimination as an explicit category of analysis when assessing racial and ethnic minorities wellbeing. Thus, the analysis of racial disparities in Colombia requires a multidimensional framework to evaluate the impact that social categoriza tion based on racial distinctiveness has in furthering inequality and impoverishing Afro-d escendants. From this perspective, this paper supports the view that Amartya Sens Capability Approach constitutes an appropriate framework for such an evaluati on as it allows for the assessment of the impact that race-based discrimination has over the real freedom an individual enjoys, and in consequence over his/her achieved welfare. Furthermore, this paper agrees, as do other authors (Kuklys 2005; Di Tommaso 2007; Krishnakumar and Ballon 2008), with the idea that latent variable modeling is an appropr iate method when assessing Capabilities and Functionings using statistical da ta such as national surveys. Finally, from the perspective of policy, this thesis conclude s that increasing direct political par ticipation, effective enforcement of the rule of law, and affirmative action policies aimed at improving access to education, health and housing, among other things, coul d constitute an effective mechanism to expand the capabilities of t hose of African descent in Colombia. 4
To provide a clear analysis of the Afro-Colombian predicament, this thesis is organized into five parts. The second chapter comprises descriptive statistics of the living conditions of the Afro -Colombian population in rela tion to non Afro-Colombians: an analysis that gives a panoramic picture of the issue at hand. The third chapter presents a literature review on social stratificati on theory. I seek to demonstrate that this theoretical framework falls shor t when assessing racial dispar ities and that Amartya Sens Capability Approach allows a more accurate depiction of these harsh inequalities. The fourth chapter illustrates the formal MIMIC modeling and its logic, and also provides a more explicit justification of its use for this study. The fifth chapte r presents results and findings. Finally the sixth chap ter briefly concludes the en tire paper, and the appendix presents a map with the geographical loca tion of Afro-descendants in Colombia. 5
Chapter 2. Background: Afro-Colombian Living Conditions Colombia in the wider context of Latin America and the Caribbean has been one of the nations living under the so called myth of racial democracy. This myth is based on the belief that Latin American societies are the product of a harmonious co-existence and mixing of the different racial groups, and th erefore these are societies free from racial tensions. The ideal of societies free from racial divisions emerged as a counter discourse against colonial powers in Latin American in the late 1800s. Latin American elites saw in the development of an all inclusive nationa l discourse the power to create compliant societies to forge modern nations. However, as national elites attempt to bring modernity, meaning imposing European models into Latin American nations, they soon realized that their racial composition presente d a threat to such process as they saw themselves and were seen as inferior by th e white power nations, includi ng the Unites States. Hence, Latin American intellectuals increasingly aw are of the international exclusion their nations suffered as a consequence of thei r ill racial composition saw to develop contesting discourses to separate themselves from traditional conceptions of race and nation. It is by 1925 that the Mexican minister of education published a remarkable essay called, the cosmic race, in which he praises the mixing of races in Latin America as an evolutionary process that had eventually lead to the born of a new race with the best of all races, the cosmic race (Miller 2006). By 1933 another influential work would be published following this same reasoning, Casa-Grande & Senzala wrote by a Brazilian 6
cultural anthropologist Gilberto Freyre. This author went beyond the idea of exalting the mixed racial composition of the Brazilian society toward the initial delineation of the myth of racial democracy. The author argues that race was not a categorical division of the Brazilian society given the almost pate rnalistic relationships established between black and whites since slavery times. Furtherm ore, in developing the conceptual content of the myth, the author argues that class and not race was the re ason for the poverty and systematical exclusion blacks and indige nous peoples suffered (Skidmore 1993; Reiter and Mitchell 2009). Colombia was not strange to these proces ses of national and racial identities formation. Furthermore, Colombian elites con front the same exclusionary treatment in the international arena as a consequence of th eir so called racial inferiority. In this scenario, and given the geographical segrega tion of blacks and indigenous communities in the country; after the wars of independence, national identity was also forged according to the cult to the mestizo nation or the cult to the mixed race that had the best of all races (Lasso 2006). The cult to the me stizo nation led to development of a national discourse based on the denial of racial categor ies as the source of social conflict. Public discussions of the topic of race and raci al discrimination were forbidden, and any mentioning of it was considered unpatriotic. As the nation was consolidating itself as a modern democratic society, the denial of r ace was increasingly support in ideas of racial democracy. Nowadays, the myth of racial de mocracy survives and the Colombian society is still living at great extent in a racial denial (Wade 1993). Nevertheless, as many historians and anth ropologists have point ed out, there is neither a complete blend of races nor societies free from racial tensions. Racial 7
democracy may be just an illusion but it is a powerful one (Lasso 2007; Andrews 2004). Its strength can be recognized in that it has served the pu rpose of purging the topic of race and racial discrimination from political debates. The recent attention directed towards race and racial discrimination has taken place mostly due to international pressures. In this scenario, racial di scrimination has never been acknowledged in Colombia; so the precarious conditions, in which ethnic minorities live, have not been linked to its existence and pervasiveness. These precarious conditions have been historically explained mostly by class differences, instead of racial ones, which are explained through a sequence of lower achieveme nt in educational attainment and labor status that in turn determine lower income levels for such individuals. In the midst of this denial, Afro-descendants, and ethnic minorit ies in general, in Colombia are the poor among the poorest. They have been historica lly marginalized and hold lower standards of living in almost all dimensions of human life. Although, there are several possible causes for these disparities, such as a cumulative result of lower income and educational attainment, the persistence and magnitude of these inequalities suggest that there are embedded practices of racial discrimination that are a disadvantage to these subgroups of the Colombian population. The first event that suggests the existen ce of racially discri minating practices in Colombia refers to the fact that official st atistical data collected in the country have systematically neglected ethnic and racial minorities, including the Afro-Colombian population. Although, this is not an exclusive Colombian phenomenon in the broader set of Latin American countries with ethno-raci al minorities; it does raise concerns, since Colombia is home of the second largest black population in Latin Am erica, after Brazil. 8
For instance, according to 1993 census data, the Afro-Colombian population was around 1.5% of the total population. Later on, accordi ng to the 2000 National Household Survey, this population increased to around 17.9% of the total population. More recently, Urrea argued that Afro-Colombians represent between 20% and 22 % of the national population: between 8.6 and 9.5 million people (2006). This last statistical account has been widely accepted by the academic community including the Economic Committee for Latin American and the Caribbean (ECLAC), since the data was assessed after using different methodologies and surveys. However, it can be observed that the 2005 Census a total of Afro-Colombian population of 10.3% of the total population (s ee table 1 below), which is not coherent with the previous number given for 2000 based on the National Household Survey. On this regard, Urrea (2006) explai ns how different methodologies and survey questionnaires lead to differe nt numbers. Accordingly, vari ations in data, instead of reflecting real demographic changes, must rath er be attributed to unreliable and changing census tools, as well as to widespread tendenc ies to whiten census participants and thus contribute to their invi sibility. This invisibility then reinforces the marginalization and exclusion of the Afro-Colombian population. Governmental recogn ition of their large presence and the discrimination they routin ely suffer is a more recent event (Arocha 1998; Urrea 2006; Escobar 2003). Geographically, Afro-Colombian settlements are located mainly in isolated zones of the Pacific and Atlantic coasts of the country.2 The Pacific and the Atlantic coasts are the regions with the largest population of Afri can descent, which by contrast are also the least developed regions of the country. Although these areas are rich in bio-diversity, this 2. See map on appendix A. 9
richness has become a threat, because it le d to the expansion of licit and illicit monocrops such coca and oil palm, often controlled by guerillas or Para-military groups, that degrade both the soil and water sources, and ot her activities such as logging and mining. These actions often disrupt the natural balance of the territori es and increase the risk of natural disasters (Observatory on Racial Discrimination 2008). The chart below shows the distributi on of the Colombian population by ethnic characterization and urban/rural locati on based on the 2005 Colombian Census. As mentioned above, the total Afro-Colombian popul ation reported by the Census states the percentage as 10.3% out of which 7.4% is urban population and 2.81% is rural population. Table 1 Distribution of the Colombian Population by Area and Ethno-Racial Characterization %#%#%# Non Afro-Colombian 68.5%28,414,13721.2%879544489.7%37,209,581 Afro-Colombian 7.4%3,085,2482.81%1,165,26210.3%4,250,509 Total 76.0%31,499,384 24.0%9,960,706100.0%41,468,384*Based on 2005 Census Data from: http://www.dane.gov.co/censo/Total Area Urban Rural Nonetheless, it has to be clarified that the National Census uses a selfidentification question, which has been severely critic ized for its lack of reliability given the pressure that societal conditions imposed on the percep tion of individuals (National Research Council of the National Academies 2004). In the case of Colombia, this is especially important given a so cial tradition that has historically prized the white race and its cultural expressions. B lackness and Whiteness are two extremes of a line determining the lower and upper categories of the social continuum. In this scenario, 10
many people of African descent and Indige nous individuals chose not to identify themselves as members of ethnic and racial mi norities, especially in major cities where the strength of the cultural pressure becomes unbearable. As a conseque nce, official data becomes unreliable, although necessary as there are very few sources of such statistics. Hence, it can be concluded that the consistent efforts of rendering minorities invisible is in itself a strong indicato r for the pre valence of racial discrimination. The second even that suggest the exis tence of race-based discrimination in Colombia is the precarious conditions in which Afro-descendants live. Traditional poverty and inequality indicators show huge gaps between the Afro-descendant and non Afro-descendant population. This is aggravated by the spatial distribution of the Afropopulation, which as mentioned above is mainly located in isolated zones completely vulnerable not only for their l ack of infrastructure, but also for the negative impact generated by the presence of illegally armed groups. Moreover, recent waves of migration to cities like Cali, Medelln and Bogot are the result of forced displacement and extreme conditions of poverty. According to the Consultation on Human Rights and Displacement (CODHES 2008), a Colombian non-profit organization, Afro-Colombians represent about 22.5% of the total displaced population in the country. Thus, for AfroColombians, displacement is directly affec ting not only their livelihoods but also their ethnic and cultural survival. Forced displacem ent in the country is generating a spatial redistribution of poverty, as already vulnera ble farmers are being violently separated from their traditional liveli hoods and must migrate to the urban centers, engorging previously existent urban peripheries of exclusion and misery. 11
The next two tables present the percentage of population under the poverty line based on the 2003 Quality of Life Survey. Tabl e 2 uses a poverty line of two dollars per capita per day and a indigent line of one dollar per capita per day, which are the thresholds internationally defined by th e World Bank. Table 3 uses the poverty line defined by the Colombian National Department of Planning (DNP), which is higher than the one defined by the World Bank. Table 2 Percentages of Poor and Indigent Populations for 2003, by Urban-Rural Areas* Urban RuralUrban Rural Afro-Colombians43,0% 76,8%16,7% 46,9% Indigenous/Gypsies52,6% 81,0%22,0% 44,2% Mestizos/Whites36,6% 72,6%13,6% 39,2 % National Total37,3% 73,2%13,9% 40,1% Population under PLPopulation under IL National Population (Urrea and Viafara 2007, 60) based on 2003 QLS Table 3 Percentages of Poor Populations by Urban/Rural and Ethnic Membership base on the QLS 2003 UrbanRuralTotal Afro-Descendants54.9%73.6%61.2% Indigenous 59%72.3%66.5% Non-Ethnic 48.7%70%54.1%*LP=281.479,80National Population Population Under PL**t-test for sample means differential shows that the mean for the Afro-Colombians is significantly different to that of the nonethnic population group significant at the 0.01 alpha level, still they appear to have the same variance. It is possible to observe that Afro-descendants and i ndigenous communities are always at the bottom. However, neither the poverty line defined by the World Bank nor 12
the one defined by the DNP differentiate be tween rural and urban populations and the specific consumptions necessities they have. Therefore, one must be very careful when analyzing the decreasing per centage of poor people in rura l populations when going from table 2 to table 3 with 76.8% and 73.6% respectively. For instance, the threshold of the Colombian poverty line puts more weight on food expenditures in comparison to shelter or clothing; as a consequence, poverty lin es are higher for urba n populations as these populations spend more of their income on f ood compared to other items. On the other hand, rural populations have somewhat different n ecessities as they tend to be a sort of autarchy in terms of their ow n consumption (they eat what they grow), and spend their cash income in aspects such basi c services or transportation. In terms of health and education, the numbers are not more encouraging. According to the 2005 census, the infant morta lity rate among Afro-descendants is nearly twice as high as that of the rest of population: 48.1% and 26.9% per 1000 live births respectively. In addition, the infant mortality rate increases up to 77.5% in the department of Choc, which is home of about 44.1% of the total Afro-Colombian population of the country3, and where Afro-Colombians made fo r 89% of its population. The infant mortality rate in Choc is comparable to that of The Republic of Congo, Ethiopia and Mauritania in Africa, with about 79%, 75% and 75% respectively. Life expectancy for Afro-Colombians is about 66.4 years while fo r the rest of the population it is around 72.8 years It also worthy to note that life expectancy is even lower for the Afro-descendants living in the Choc department, which has c onsistently lower standards in comparative 3. Long Term Plan for the Black Population, Afro-Col ombians, Palenque y Raizal: Proposal y Process 2005 2007, National Department of Planning, 2007. 13
terms, not only with the rest of the popul ation, but also with the Afro-Colombian population located in different regions of the country. The next table shows educational data, which is strongly correlated with labor status and social mobility4. The data present Afro-Colombians behind non AfroColombians consistently, and it is especially important to note that college education, which is considered one of the most importa nt contributors to rising out of poverty, is almost three points lower for Afro-descendants in comparative terms with non Afrodescendants with 5.21% and 8.07% respectively. Table 4 Educational Level by Ethnic Characterization based on the 2003 Quality of Life Survey #%#%#% None 392,79012.90%2,916,1908.19%3,308,9808.56% Elementary 1,275,56641.88%14,460,00040.62%15,735,56640.72% High-School 1,101,04436.15%13,400,00037.64%14,501,04437.52% Technical 84,5272.78%1,387,0423.90%1,471,5693.81% College 158,7885.21%2,872,5358.07%3,031,3237.84% Graduate 33,0311.08%563,9661.58%596,9971.54% Total 3,045,746100%35,599,733100%38,645,479100% *It excludes indigenous population, which is around 2.16% of the population or 942.115 individuals as estimated by the ECV 2003. **An equality of variance test was perfomed yielding a Folded F-statistics of 1.35 woth a p-value of <0.0001. Thus, we conclude that the variances of the two subgroups are different. ***The t-test result was equally significant at the 0.01 critical level, hence we conclude that the means for these two groups of the population are in fact different. Educational Level Variable Afro-ColombiansNon Afro-Colombians* Total Total Total 4. In Colombia development gaps between urban and rural populations are very high. The fact that the 2003 Quality of Life Survey under represents rural population might cause an overestimation of the e ducational level for ethnic minorities, indigenous and Afro-descendants, which are main ly rural. Especially the numbers for college and postgraduate education may be evidence of such underestimation. 14
In addition, according to the 2005 Census data, 11% of Afro-Colombian children did not attend elementary school, while for hi gh school education that percentage went up to 27%. The reasons for school absenteeism ra nge from the lack of money to cover the cost of tuition and materials, to the lack of infrastructure, and to the availability of teachers. These problems persist despite the States obligation to provide free elementary education up to the fifth grade to the entire population. The inability of the Colombian State to guarantee basic rights to its ethnic minorities contrast s sharply with the countrys laws and human rights protocols sign ed at the international level. In 1991, Colombia promulgated a new National Constitution5 in which ethnic groups and minorities were not only granted special rights as compensation for a history of marginalization, but also as a means of preserving ethnic communities within their own cultural, social and economic structures. In addition to the Constitutional articles, there are three additional legal laws and norms directed towards fighting discrimination, yet the judicial system lacks effective guidelines to comply with these laws6. As a result, neither racial discriminators are prosecuted nor are victims redressed. Finally, there are no preventive measures to avert future episode s of racial discrimination (Observatory on Racial Discrimination 2008). Moreover, neither the adoption of the International Labor Organization (ILO) convention about Indigenous and tribal communities7 nor the signing of the International Convention on the Elimination of all Forms of Racial Discrimination 5. Afro-Colombians were granted with collective property rights, community-orient development and prior consultation rights, and the right to be free of discrimination in the transitory article 55, later developed in the Law 70, 1993. Comisin Consultiva de Alto Nivel para las Comunidade s Negras Subcomisin de Planeacin y Desarrollo Ley 70 de 1993 Decreto 2248 de 1995. 6. Article 13 of the national Constitution, articles 58 and 147 of the penal code, article 48 of disciplinary code for official functionaries, and article 33 of Law 70 of 1993, were introduce into the judicial system to punish acts of discrimination based on race or ethnicity. However as argue by the Observ atory on Racial Discrimination (2008, 60) none of them are effectively enforced. 7. It was integrated into the national legislation by the Law 21 of 1991 (Ley 70 de 1993 Decreto 2248 de 1995). 15
of the United Nations8 have led to any major improvement in the living conditions of the Afro-Colombian population, which points to a weak state, ra ther unable or unwilling to protect all of its citizens and extent citizenship rights to them. Nowadays, after more than 150 years after the abolition of slavery, 18 years after the enactment of the 1991 National Constitution and the signing of several international treaties and protocols, Af ro-Colombians are still persistently marginalized. Unfortunately, legal recognition of Afro-Colombi an rights has not been translated into an effective enforcement mechanism and public policies directed to address the marginalization of Afro-Colombians. This l ack of a political and judicial mechanism to guarantee basic and constitutional rights for et hnic minorities in conjunction with the harsh reality of exclusion of Afro-Colombians, points toward the proposition that their marginalization is less the result of soci al stratification and more one of racial discrimination. This is not to deny that clas s disparities and inequalities in access to resources have an important share in determin ing the lower standards of living of those of African descent; what is really argued here is that racial discrimination is an important underlying cause of these preca rious conditions as well. To summarize, Colombia is a society in which the historical and social construction of racial identities assigns lower values to indi viduals of black and indigenous backgrounds. This has its expression in lower st andards of living for AfroColombians. They still have limited access to education and health care facilities even after legislation to change these conditions. They have higher mortality rates, lower life expectancy. Their cultural and territorial rights are constant ly threatened by mainstream 8. New York, March 7, 1966, and ratified in 2004. 16
development models based upon the over-explo itation of bio-diversit y; and by the large presence of armed groups such as guerrilla s and paramilitaries. Their vulnerability is endemic to all aspects of human life since th ey are socially, economic ally, politically and culturally excluded from Colombian societythe y lived as second class citizens. In this scenario, the need to go beyond the myth of racial harmony and to question the explanatory power of traditional analysis of class disparities is evident. 17
Chapter 3. Theory and Model Building Race, Racial Discrimination and Poverty Group disparities have been increasingly studied across academic disciplines and from different perspectives. These studies have been motivated by the existence of pervasive inequalities across racial and ethnic groups around the world (National Research Council of the Nati onal Academies 2004). From a sociological perspective, social stratification theory has been the main analytical framework to analyze the impact that group disparities have on individual welf are. This theoretical framework assumes class as the main category of analysis, and uses the educational achievement and employment status of an individual as a tool to understand the dynamic elements that interact in a process of so cial stratification and mob ility (Grusky and Kanbur 2006). These processes are usually divided between social and individual paths of mobility. Mobility in the social space is determined by the extent to which children are able to attain higher educational achievement and occupa tional status in relation to their parents. Individual mobility is determined more upon pe rsonal characteristics such as gender and race (Moynihan Ed. 1969). A democratic society would then be characterized by individual and social patterns of mobility that do not depend upon th e social class of their parents and the ascribe characteristic of an individual, but rather on his/her merits based on his/her 18
capacity. Equal opportunity access should be a precondition for an equality-based democracy, in which different strata are dete rmined by ones personal abilities. In the absence of such an ideal reality, academics have concentrated their analyses on the impact that class disparities have over the availability and distribution of market-based opportunities for upward social mobility across groups. Analyses of group disparities (race, ethnic ity and gender) in Colombia are mainly concentrated in social mobility and stratific ation theory. There are a number of studies such as those of Medina (2002), Portilla (2003), Viafara (2005) that analyze group disparities in education and th e labor market, and how those impede processes of social mobility for historically disadvantaged groups such as racial and ethnic minorities. Medina (2002) analyzed the differences in the labor supply according to the racial background and gender of the individual, specifically women. He divided the labor supply of women into three categories, inde pendent, wage earner and unemployed, and found that the women of African descendent ha ve less time to offer in a working capacity because they have on average more dependent children. Thus, lower labor supply is associated with higher dependency rates (chi ldren) and poverty is a likely outcome of such a combination. Nuez and Ramirez ( 2002) found that at the micro-level the educational level in the household and the num ber of children have a significant impact on poverty levels9. While a higher educational level de creases the probabilities of being poor, a higher number of dependable children increase it. Hence, people of African descent are more likely to be in poverty a nd located in the lower strata of the class continuum because they have lower educa tional attainment, lower labor status, and 9. This study is concentrat ed on macro-level determinants of poverty, but it includes micro-level factors as they identified causal mechanism from economical aggregates at the m acroeconomic level toward economical variables at the microeconomic level. 19
through these, lower income. Also, Viafara (2005) concluded th at in the city of Cali, Colombia, an individual member of a house hold of African descent with a given socioeconomic status (measured as the educational a ttainment of the head of the household) is more likely to achieve lower occupational stat us in comparative terms with white or mestizo individuals with the same socio-ec onomic status. According to the author, this illustrates how processes of social mobility ar e influenced by ascriptive characteristics of the individual and not just by income levels. Unfortunately, social mobility theories present different shortcomings that should be kept in mind. First, these theories clas sified people according to their social class, which in turn is determined by the intera ction of their achieved schooling, occupational status and income level, in that order, as one causes the other (Grusky and Kanbur 2006). This categorization represents an important flaw when analyzing poverty and inequality given that once an individual moves up in the so cial strata, which is strictly associated with his/her economic wellbeing; the persistent inequalities associated with race follow them. This means that monetary inequality is reduced and the individual is no longer considered poor or part of the objective population of welfare polic y; however, in the case of racial inequalities, racial discrimination does not end once people from ethnic minorities move up in the social strata. On th e contrary, even people that are over these income thresholds have to endure discrimina ting practices across different dimensions of human existence. This is very critical in a country like Colombia where the rule of law is not effectively enforced and racial discrimination is not addressed by judicial mechanism. Thus, people suffering the effects of racial discrimination do not have the means to defend their rights even when they are not considered poor by standard measures. 20
Second, even when these studies do analy ze the role of ascriptive characteristics such as gender or race; concluding that these aspects played an important negative role in the stratification process and the distribu tion of opportunities for upward mobility among social and ethnic groups; their exclusive focus on resource access as the major source of inequality and poverty reduces th e scope of the analysis they can make regarding racial inequalities and discrimina tion (Blau and Duncan 1967; Schiller 1971). These theories share core aspects with traditi onal analysis of pove rty and inequality in the income space severely criticized by several authors, including Atkinson (1970), Rawls (1971) and Sen (1980, 1985, 1997, 1999). Sen argues specifically in his criticism that the value attached to a given dimension (education and occupation) is supported by its instrumental value as a means of increasing wealth and income. Th erefore, aspects such as empowerment, the command of social respect, livi ng without shame, and the fulfillment of ones rights on the basis of ones humanity and citizenship, ar e not part of the informational basis of these frameworks. As a result, intrinsic values and rights are not considered by themselves, but as mere instruments in the ach ievement of wellbeing; more exactly in the achievement of economic wellbeing. Howe ver, as Amartya Sen (1985, 1997, 1999) has shown, economic wellbeing is just a component of human wellbeing, an important one but nonetheless just a component. Furthermore, the possibility to live a life free from discrimination is a core aspect of human wellbeing, especially for those groups who have had to endure it from historical times to the present and ar e still suffering the consequences. Such consequences cannot be completely accounted for by analyzing economic wellbeing. Human wellbeing goes well beyond the mere availability of economic resources into the realm of intrin sic freedoms. After all, to determine the 21
quality of life of an individual or group, we have to know not just how much income a person or group has, but how much of that in come contributes to th eir ability to live a decent life. For these reasons, this thesis argues that the need to go beyond traditional analyses of group disparities and social mob ility theories is unmistakable. The Capability Approach and its human centere d perspective allow one to overcome the shortcomings of social stratification theory with respect to th e analysis of racial inequalities. Given its multidimensional nature, Sens theoretical framework can account for the impact that discrimination patterns have over the wellb eing of individual members of racial minorities beyond income levels. Amartya Sens Capability Approach Amartya Sens Capability Approach is centered on the enhan cement of individual freedoms as both the end and means of deve lopment. According to Sen, any welfare assessment must account for the process of expanding or constraining individual freedoms. The Capability Approach is based upon the recognition of the multidimensionality of both welfare and poverty, as these are conditions determined by a wide range of factors. He argues that the ev aluation of welfare must be examined with respect to the real freedoms people enjoy, and not just with respect to the income they have. The author criticizes standard welfare economic approaches that concentrate their analyses in terms of we alth (opulence, income and/or expe nditure) or utility (happiness or human fulfillment). He argues that the analysis in terms of opulence assumes that income 22
is an end in itself; conse quently, it does not account for individual differences when converting income into valuable achievements (Sen, 1985a; 1985b; 1999). Furthermore, the relation between low in come and poverty is instrumental and variable. Instrumental in the sense that there are other fact ors that influence poverty and income is just one means among others to re duce it. The ability to convert income into better living standards varies greatly among communities and individuals. In addition, as was noted also by Rawls (1971), the analysis of individual welfare in terms of utility does not set apart different sources of satisfaction and distaste, and cannot reflect the intrinsic value of aspects such as human rights. Besides, these traditional frameworks assume that individual rationality is base d on self-interest, an assu mption that has been widely criticized as people often consider wide r concerns when deciding among a set of alternatives (Sen 1985b, 1997, 1999). Sen states that neither utility nor income allows us to evaluate comprehensibly individual welfare and social development. To fill these theoretical gaps, Sen developed the Capability Approach, which assesses individual wellbeing by considering re alized welfare and the freedom to choose among different options in order to reach fulfillment. It identifies two elements, functionings and capabilities and their corresponding sets. Accordingly, functioning analysis in the Capability Approach evaluates individually achieved wellbeing by considering a wide range of aspects such as education, employmen t, health, nutrition, social or political participa tion, and the way in which these factors interact. Capabilities, in turn, represent the set of op tions available to the individu al to achieve his/her desired level of wellbeing (Alkire 2005; Saith 2001). Se n defines individual agency as what the person is free to do and achieve in pursuit of whatever goals or values he or she regards 23
as important (Sen 1985a, 203); and in his book Development as Freedom he argues that the aim of the development must be the expansion of individual freedoms (1999). By this definition, the capability set is determined by the extent to which an individual can exercise his/he r agency or individual freedom and the aim of development is to expand said agency. However, not just any increase in the capability set matter or count as an expansion of freed om. In order to be able to achieve a real or effective increase of capabilities, it has to be valued by the individual or group as an increase in their set of choices; otherwis e there is no capability e xpansion even though there are more choices (Alkire, 2005). Hence, indi viduals must be agents of their own development as they are the ones who must reas on and decide what constitutes a valuable end in their lives. Thus, the Capability Appr oach praises the role that active citizen participation has, as a precondition for the successful development of policy design and implementation. A clear example of this s ituation is found in the Afro-descendants communities that live in the Atrato river valley in the department of Choco, Colombia. These communities have collective property and prior consultation rights, in addition to the right to chose development projects that are in accordance with their cultural and ethnic survival preferences. However, such as rights are not enforced and these communities in addition to suffer the pressure of violent displacement as a consequence of the armed conflict, are also sufferi ng political pressure to accept unwanted development efforts. Effort such as the harvesting of m ono-crops and the over exploitation of the rich biodiversity of the region among other economic initiatives are unwelcomed initiatives for these communities. As a consequence, even though these are efforts labeled under development initiatives that eventually coul d improve the economic 24
wellbeing of these communities; they do not represent an expansion of capabilities as these initiatives are actually s een as a threat to their cultural surviv al (Escobar 2005). Sen (1999, 10) does not dismiss the instrume ntal role of different freedoms such as (1) political freedom, (2) Economic fa cilities, (3) social opportunities, (4) transparency guaranties and, (5) protective security. ; on the contrary, he explains that they interplay a major role in enhancing i ndividual capabilities as they reinforce one another.10 But Sen goes beyond the widely recognized instrumentality of the different dimensions of human welfare, by arguing that these dimensions also have an intrinsic value, as they are central as pects to the process of human flourishing (Sen 1999; Alkire 2005). Sens Capability Approach considers circumstances su ch as poverty, political repression, racial and gender discrimination, unemploym ent, lack of economic, educational or health facilities as capabil ity deprivation. Therefor e, from a capability perspective, the persistent de privation of Afro-Colombians would be as much the result of insidious patterns of racial discrimination as it would be the accumulated result of low income and class disparities. In Sens arguments, what matters is the freedom or the effective capability to achieve valuable f unctionings. These capabi lities are determined by the freedom a person enjoys in achieving the life she or he has reason to value, which alludes to the fact that different individua ls living in different societies, will have different demands or necessities. In this lin e of reasoning, income is just one means among others used to expand one s capabilities and to achi eve valuable functionings. Thus, low income is just one reason among many to explain the pe rsistence of group 10. This interconnectedness between the differe nt dimensions of human existence, welfare is also developed as core aspects of the Human Security Approach (Nef 1999). 25
disparities. The achievement of valuable beings and doings is highly constrained by factors such as gender, race and ethnic backgrounds, social arrangements as well as disabilities and environmental constraints; moreover, these factors affect greatly the ability to translate income into valuable achievements (Sen 1999, 88). Consequently, when considering a wider range of the as pects influencing hum an wellbeing, class disparities and income inequality analyses fe ll short, especially in the case of racial minorities subject to discrimination. These shortcomings can be overcome by analyzing capabilities and functionings as their multidimensional nature allows us to include previously excluded factors when asse ssing Afro-descendants welfare. In this case, the Capability Approach allo ws us to analyze the role of race as a socially imposed handicap that reduces the ca pabilities or freedom of those of African descent. Under the Capability Approach, considering the role that racial discrimination plays in limiting the exercise of individual freedoms for ethnic and racial minorities must be an explicit category of anal ysis. If race-based discrimination is included as a specific category in the assessment of individual welfare of Afro-descendants, there would probably be a significant increase in poverty levels as measured by this multidimensional approach in contrast to that of the income poverty measures that disregard race. This is explained by the fact that the Capability Approach evaluates more than income when judging welfare conditions. Thus, from a capab ility perspective, individuals over the poverty line, even located in the middle position of the inco me deciles, could fall in poverty as they are the subjec ts of insidious patterns of racial discrimination. Such negative impact cannot be asse ssed by income level analysis, which in turn, corroborates the poorness of income-measure based approach es when assessing more subjective but 26
equally important aspects of human welf are (e.g. empowerment and the ability to command social respect and not just commodities). Notwithstanding its increasing importan ce and relevance, the Capability Approach has been severely criticized due to its difficulty when setting it into operation. Some critics have reconstrued its main arguments as theoretical weaknesses. For instance, because the Capability Approach expands the informational basis when evaluating individual welfare, this has been reevaluated as a disadvantage given the difficulty in elaborating on broad theoretical concepts with equa lly broad empirical significance (Clark 2006; Commin 2001). Furthermore, Martha Nussbaum (2000, 2003) and Severine Deneulin (2004) have labeled Se ns approach as incomplete since it does not claim to be a theory of the good, and he nce it does not set normative parameters to guide the enhancement of individual free doms in a specific society. Nussbaum (2000) drew a set of central human capabilities with a heavy Aristotelian influence in order to give Sens approach the completeness it was lacking. However, Deneulin (2004) argues that although Nussbaums pragmatic appr oach achieves the setting of normative boundaries to guide developmental efforts, both Sen and Nussbaums approaches fall short when determining the mechanism through which individual freedom is to be enhanced. According to Deneulin (2004), Sens Capability Approach must include in its informational basis what she calls socio-historical agency or the socio historical factors affecting individual freedoms in every specifi c society, and which cannot be reduced to simply individual decisions and actions. Still, Sen (1997, 1999) argues that the evaluation of social arrangement should be made by the fi nal impact these have over an individual, 27
thus, collectivity is assessed whenever it a ffects individual freedom because is at the individual level that the change matters. Despite these critiques, the Capability A pproach has constituted itself as a leading theoretical framework when analyzing indivi dual welfare. Author s such Alkire (2005), Kuklys (2005), Krishnakumar and Ballon (2008) and Di Tommaso (2007) found the approach useful, and identified different methods to implement it. Alkire (2005) evaluates the expansion of capabilities among Pakistani wo men part of an (OXFAM) program directed to increase the particip ation of women in income earning and empowerment. She combined economic analysis of cost-benefit with qualitative analysis of participatory developmental policies, in or der to define and eval uate the expansion of capabilities. She used both microeconomic and qualitative data to assess the real expansion of capabilities. However, qualitative data at the micro level presents critical problems when generalizing findings given its us ually reduced scope, an d the fact that the researcher is the gatherer of the data, and t hus making it more subjective. In addition, this type of data is more detailed and its collect ion is more time consuming. Hence, the cost of collection, analysis and re porting of findings is higher a nd this represents a giant obstacle in having it collected in a systematical way, given the scarcity of funds for such enterprise.11 As a consequence, there has been no t only an increasing awareness of the need to collect capability data in a systematical way, but also the need to use existing data to analyze capabilities (Seth 2009). The use of the latent variable method to operationalize the Ca pability Approach has steadily gained prominence in its applic ation. Several authors have applied this 11. John Gerrring, Case Study Research (Cambridge: University Press, 2007). 28
methodological approach when implementing the Capability Approach with significant results. For instance, Kuklys (2005) used latent va riable modeling to evaluate the welfare conditions of the disabled population in England. She makes an inter-temporal comparison of Sens functioning achievements in health and housing between 1991 and 2000. Her work concluded that disabled pe ople in England are even poorer when analyzed under the Capability Approach, given their pers onal handicaps when converting income into valuable achievements. In a ddition, Di Tommaso ( 2007) assesses childrens welfare in India using longitudinal data a nd structural equation modeling. Her document identifies three capabilities: bodily hea lth, senses imagination and thoughts, and leisure activities and play as key aspect s of childrens welfare. The study found that among other factors, gender is a key aspect in childrens welfare since boys have a higher possibility to access education. Also, Kris hnakumar and Ballon (2008) applied the same methodology to Bolivia, identifying two key capab ilities as latent constructs: Knowledge (being able to be educated), and living conditi ons (being able to be adequately sheltered). They use variables such as years of edu cation and dwelling conditions of the house as indicators of the knowledge and shelter ca pabilities resp ectively. In addition, they hypothesized causal relationships between a se t of exogenous variables such ethnicity, gender and age and the latent capabilities. Among their findings, this study found that belonging to an indigenous gr oup played a key role in dete rmining childrens welfare in Bolivia, having a consistently negative impact. In conclusion, several authors mentioned have applied the Capability Approach, using latent variable as empirical method w ith significant results to their specific cases. They have shown that given the complex nature of Capabilities, its empirical application 29
requires a methodological approach flexible enough to capture their complexity, and latent modeling has shown itself a useful method for such a task. Its methodological refinement allows for the capturing of the complex nature of capabilities, using common statistical data, such as national census, which in turn are more reli able, less costly and widely available. Thus, the de cision to use MIMIC modeling w ith existing data to apply the Capability Approach to the case of Colomb ia has some practical reasoning behind it. These reasons are the availability of data and the exploratory nature of this study, as there almost no studies that analyze racial inequalities in Colombia from a multidimensional perspective. 30
Chapter 4. Operationaliz ing Amartya Sens Capab ility Approach: A MIMIC Model Applied to Colombia Data and Conceptual Clarifications The data used to estimate the Multiple Indicators Multiple Causes (MIMIC) model comes from the 2003 Quality of Life Survey (ECV 2003 in Spanish) executed by the National Department of Statistics of Colombia (DANE in Spanish). The survey defined nine geographical re gions: Antioquia, Atlntica, Valle, Pacifica, Central, Oriental, Bogota D.C., Orinoquia-Amazonia, San Andres y Providencia It is based on direct interviews with all people older than 18 years in a household, and it is representative at the nati onal level with a sample of 24,090 households and 85,145 individuals. However, the survey undervalued the Afro-Colombian population for two reasons though it is more probl ematic with the indigenous population. One is that the ethnic and race variables were taken as equiva lents. The other is that Amazonas, Arauca, Caquet, Casanare, Guaina, Putumayo, Vaups y Vichada were all regions in which rural areas were not intervie wed; so there is an undervaluation in terms of rural population in general, and of ethnic minorities specifically since these are mainly rural communities. Nonetheless, it has been widely used for its completeness regarding information about living conditions and its representa tiveness at the national level. It is necessary to clarify that although the selection of the different capabilities is supported by different theo retical approaches and previo us studies, there are other 31
capability dimensions that cannot be analyzed given data constraints. For example, in the 2003 Quality of Life Survey there is no info rmation regarding political and community participation, which are factors widely r ecognized as important dimension of human welfare. In addition, the sample is not repr esentative at the state level, thus it is impossible to make analysis and comparisons with the department of Choc. This could constitute a rich scholarly exercise that could allows to make inferences including variables of geographical segr egation since about 45% of the Afro-descendant population is located in this zone. Also, it is important to note that there are intrinsic limitations when assessing racial discrimination by fitting a statistical model to available data. For instance, omission of relevant variables and sample bias ma y occur making the inference of race-based discrimination controversial.12 However, this type of data provides reliable information on different outcomes by groups, th at together with a careful contextual and statistical analysis, allow the inference of possible causa l factors and mechanisms that disadvantage a specific group, including the role of race. The analysis of group disparities using statistical methods and availa ble data must deduce that di scrimination is present, by determining if a specific outcome would have been different if an individual would have been from another gender, race or ethnicity. If the results show that the impact of race is still significant after the other variables have been held ceteris paribus13; there is a high probability that racial discrimination is ta king place and is disadvantageous to a given population group (National Research Counc il of the National Academies 2004, 5). 12. For a more detail discussion of thes e issues see National Research Council of the National Academies (2004) Chapter Seven. 13. Ceteris Paribus equals in this context statistical control. 32
For analytical purposes, this study uses the definition of race and racial discrimination given by the National Research Council of the Nationa l Academies in its book report Measuring Racial Discrimination In this report, race is defined as a subjective social construct based on observed or ascribed characteristics that have acquired socially significant meaning (2004, 2). Also, racial discrimination is defined in two components: (1) differen tial treatment of the basis of race that disadvantages a racial group and (2) treatment on the basis of inadequately justified factors other than race that disadvantages a ra cial group differential eff ect. (2004, 4). Finally, the calculations were performed using the Stat istical Analysis System (SAS) and its procedure Covariance Analysis of Linear Stru ctural Equation (CALIS) which is used to implemented analysis of covariance structur es, path analysis and linear structural equations. Since the majority of indicators and exogenous variables were dichotomous or categorical, the estimation method used was Weight Least Squared14, is a dist ributional, assumption free method, meaning that it doe s not assume normality of the observed variables, used to estimate the set of parame ters implied by the model. It uses a weighted sum of squatted residual matrix to mini mize the discrepancy between the sample covariance matrix, S, and the model implied covariance matrix, 15. Although latent variable modeling is bei ng increasingly used in the empirical application of the Capability Approach, it is necessary to mention that the method has its shortcomings. In first place, since it is a st atistical technique of confirmatory character the possibility to test new hypothesis about cau sal relationships is limited. This represents 14. The default weight matrix for weighted least-squares estimation is an estimate of the asymptotic covariance matrix of the sample covariance or correlation matrix. In this cas e, weighted least-squares es timation is equivalent to Browne's (1982, 1984) asymptotic distribution-free estimation. (SAS Corp. 1999). 15. In this case the weighted matrix to fit the data was about forty pages; thus for spacing a nd practical considerations it would not be presented in a paper as part of the resu lts, but the information woul d be available on request. 33
an important shortcoming in terms of the ope rationalization of the Capability Approach due to the intrinsic incompleteness of the fr amework in need of exploring new paths and hypothesis testing as the framework evolves. In second place, these methods do not allow the quantification of latent constructs; it just allows to infer its behavior from the set of observed indicators. In this case we cannot quantify a Capability Set, how many opportunities or choices a persons enjoy in a given dimension. However, this does not demerit the usefulness of the approach as it does allow the quantification of causal factors of Capability expansion or c ontraction. Finally, while the appr oach presents fine qualities regarding the applicatio n of the Capability Approach; it suffers the vast limitation of the lack of pertinent statistical data. Although this is not a specific problem of the approach, it does limit its scope of analysis. MIMIC Modeling Latent variable16 modeling is a much more robus t approach than regression, although it is a member of the General Line ar Model (GLS) family of which multiple regressions is a part. Regression analysis is a statistical method to analyze data aimed at predicting a criterion variable in terms of a set of explanatory variables. The idea is to maximize the portion of the variance of the dependent variable explained by the set if independent variables. However, while re gression cannot introduce measurement errors on individually observed variables explic itly, latent modeling allows for such approximation. In a regression model, achieved years of education would be considered an explanatory variable free from error, a nd the model would yield a composite measure 16. For more detailed information on these methods see Schumacker (2004), Bollen and Davis (2009), and Krishnakumar and Nagar (2008). 34
of error of the complete set of explanatory va riables. By contrast, la tent variable methods allow one to realize to what extent the variable years of education measures the knowledge level of an individual. Since such measurement coul d have an erro r associated with the quality of education that the individua l actually received, this method allows us to account for that error. In a ddition, this approach presents different options for handling missing data; it allows one to analyze the influence of intervening variables, and also it allows for testing for misspecification of pr oblems all in one (Schumacker 2004; Garson 2008). MIMIC models were developed by Jreskog & Goldberger (1975) and is part of this much larger family of statistical techniques called latent vari able methods, which combine path analysis and confirmatory factor analysis techniques. Path Analysis is a statistical technique use to decompose the relationships between a set of variables, which can be independent, intermediary and dependent17. It is considered an extension of the regression model because it analyzes the impact that intervening variables have in the relation between independent a nd dependent variables. Furthe rmore, in path analysis, a variable can be independent, dependent or both. Neither an intervening variable nor a variable that is both dependent and independent can occur in regression analysis (Garson 2008). Confirmatory factor analys is, in turn, is a statistical technique directed toward identifying if a set of variables, called i ndicators, to measure a latent construct or unobserved variable. In factor analysis, the idea behind a late nt construct is that its complexity makes its direct observation impossibl e, and it is necessary to rely on a set of 17. Path analysis is often called causal analysis, but this na me is explained more by the relationships hypothesized by the researcher than by the statistical outcome of the technique. This technique tests the significance of the correlation matrix of the data, given the models specified and compared by the researcher, and which must be supported on firm theoretical grounds and previous research (Schumaker 2004; Kline 2005; Garson 2008). 35
indicator variables to infer its behavior. In our case, a confirmatory factor analysis example would be just the relationship between the latent capability and its corresponding indicators, without the exogeno us causes. These two techniques, path analysis and confirmatory factor analysis, share the basic assumption that the model specified and tested by the researcher mu st be firmly grounded in theoretical and previous research assumptions (Schumaker 2004). Contrary to factor analysis, MIMIC modeling allows for the inclusion of a set of observed variables as exogenous ca uses of the latent construct. In this sense, the latent construct can be depicted as an interven ing variable between the exogenous causes and the observed indicators, as path analysis doe s but only with observed variables. Similar to path analysis and factor analysis, MI MIC modeling is confirmatory rather than exploratory as it assumes that the relations between the i ndicators, the exogenous causes, and the latent variables are supporte d by theory and previous research. Capabilities are the set of choices from which each individual chooses to achieve certain levels of welfare in each dimens ion of human life. These Capabilities are determined by individual and contextual char acteristics, and in turn, they determine the achieved welfare or deprivation of an indivi dual (functioning achievement) in the given dimension. It must be noted that different f unctioning levels can be associated with the same level of freedom or capability; as each individual may or may not choose to exercise their individual agency in achievi ng a given level of we llbeing. Nevertheless, from a capability perspective, th e fact that there is freedom to choose is what constitutes real development (Sen 1990, 1999). For instance, in the case of knowledge: two persons A and B may have the possibility to study at a public or private univ ersity to achieve a 36
professional level of education. After five ye ars A graduated and B did not. It is clear then that although both enjoy th e same capability set, their choices differ, and their level of achieved welfare or functioni ng varies too. In this case, in dividual A chose to exercise his/her freedom by graduating while indi vidual B chose a different option. From an empirical perspective, asse ssing how much freedom a person enjoys, when deciding which level of welfare or functioning achievement he or she wants, represents a major task. The Capability set is determined by the interaction of a wide range of aspects, and modeling such comple xity requires a sophisticated methodology. Latent Variable Modeling, from which Mult iple Indicators Multiple Causes (MIMIC) modeling is part, has shown its elf to be a useful tool in this matter (Kuklys 2005, Di Tommaso 2007, Krishn akumar 2007). As mentioned above, the aim of MIMIC m odeling is to estimate the relationship between a set of exogenous variables (e.g. ge nder, ethnicity, poverty, etc.) one or more unobserved variables called latent (e.g. Knowledge Capability), and one or more indicators (e.g. years of educat ion, literacy, practical educat ion, etc). Both, the exogenous causes and the set of indicators are observed variables. He nce, the set of exogenous variables influence the latent construct; and in turn, the latent vari able influences the indicators. This is accomplished by the anal ysis of covariance structures between the different variables endogenous and exogenous, and the latent construct. Thus, it is clear 37
that there is need of a high correlation betw een the variables, although not as high as to not allow the estimation of parameters18. Given its robustness in relation to regres sion, path analysis, a nd factor analysis techniques, MIMIC modeling is a suitabl e methodology to implement the Capability Approach. Specifically, the use of MIMIC modeling to estimate the Shelter and Knowledge Capabilities is justified because Capabilities are the set of choices from which an individual selects what he or she wants to achieve in a gi ven point in time. What a person manages to be or do at th at point is a functioning or achieved welfare (e.g. elementary school), but that level of ach ieved welfare is not saying much about the freedom that person enjoyed--the choices he or she had in the Capability set--when achieving a specific level of welfare in the given dimension. Individual capabilities are very complex constructs that do both influence and are influenced by a wide range of aspects. This complex nature of Capabilities makes almost impossible to observe them directly. By modeling capabi lities as unobserved constructs in a MIMIC model, it is possible to account for both the factors determ ining the set of choices or the freedoms a person enjoys, and the functioning achievement that they attained given that capability set. In this case, by assessing the Knowledge Capability as a la tent construct in a MIMIC model it is possible to a ccount for the influences that variables such as gender or ethnicity has over the Capability Set, while at the same time it is possible to identify the impact on the realized welfare of an individual given a unit cha nge in the Capability Set. 18. There is not clarity regarding the impact that the lack of independence between expl anatory variables, that is Multicollinearity, would have in the estimation of Structural Equation Models. Some authors have argued this is a robust approach in the presence of su ch, yet research on this area is still very incipient (Grewal et. Al. 2004). 38
The next table shows the diffe rent variables and dimensions used to apply to this scholarly study. It identifies the set endoge nous indicators and exogenous causes linked to the latent constructs. For instance, the set of exogenous causes (e.g. gender, ethnic membership, being poor, etc.) influence th e knowledge capability or the freedom a person enjoys when deciding which level of knowledge he or she wants to have. In turn, the latent capability is inferred by a set of indicator variables such as achieved years of education, which would reflect changes in the capability dimension. Thus, when there is a factor constraining the freedom to choose the knowledge level an individual wants to achieve, this would be reflected in lowe r outcomes in the given dimension. This proposition has an intuitive meaning since it is logical to exp ect that in a given dimension of human welfare, lower outcomes are associat ed with lower levels of freedom or less options to choose the life one wants. In addition, the two di mensions of human welfare analyzed here were chosen fo r both their intrinsic and instrumental value as they are widely recognized as important dimensi ons to promote human development and individual welfare. The knowle dge capability, for instance, re presents the possibility of being educated given certain individual and environmental characteristics such as age, gender, race, marital status and the availabil ity of institutional arrangements that allows people to increase their critical thinking abili ties. This capability is translated into the functioning space as having more years of education or more practical on the job training. Finally, being properly sheltered is not only an integral part of human wellbeing, but also an instrumental factor contributing to h ealth outcomes. Living in an overcrowded environment, a lack of drinking water, a nd unsafe construction materials are factors highly correlated with a number of health deficiencies. Furthermore, there is a high 39
40correlation between each dimension and other dimensions of human welfare not analyzed as each capability is analyzed separately. However, given the exploratory character of this study, and the lack of per tinent data these interactions are not analyzed here. This exercise constitutes a much larger and complex future study.
Table 5 Variable Description and Measurement Latent Variables or 'Capabilities' Functionings or Achieved Wellbeing (Observed Indicators) Variable Measurement Individual Characteristics (Exogenous Causes of Capabilities) Variable Measurement Household Characteristics (Exogenous Causes of Capabilities) Variable Measurement Shelter Dwelling Conditions of the Household Categorical Variable (-1=low/0= middle/1= high quality) Age Discrete Risk (Household locate in a risk zone flooding, landslides overflow, geological faults, etc) Dichotomous (1=yes/0=no) Basic Services Conditions (Electricity, water, waste management and sanitary infrastructure) Categorical Variable (-1=low/0= middle/1= high quality) Gender Dichotomous (1=Male/0=Female) Area Dichotomous (1=Rural/0=Urban) Overcrowding (More than three persons per room) Dichotomous (1=yes/0=no) Female Head of Household Dichotomous (1=yes/0=no) Famishing (Not having at least one out of three meals a day given monetary constraints) Dichotomous (1=yes/0=no) Knowledge Illiteracy Dichotomous (1=yes/0=no) Ethnicity Categorical (1=Non/2=Afro/3=Indigenous ) Crime Victim (Last twelve months) Dichotomous (1=yes/0=no) Civil Status 1=Married/2=Widowed/3=Di vorced/4=single Income Decile 1 to 10 Finished Years of Educations Ordinal Being Poor (LP) Dichotomous (1=yes/0=no) Disaster (Household victim of natural disasters) Dichotomous (1=yes/0=no) Technical education for work Dichotomous (1=yes/0=no) Working status Dichotomous (1=yes/0=no) Property of the House (own, rent, hypothec) Dichotomous (1=yes/0=no) 41
The estimation of a MIMIC model is made up of two parts: First, the measurement model, which links each Cap ability (Knowledge and Shelter) to its corresponding set of indicators or observed va riables (e.g. years of education for the Knowledge Capability). This relationship is explained by factor loadings or validity coefficients that tell the extent to which a given indicator is able to measure the corresponding latent vari able. Factor loadings have a sim ilar interpretation to that of a beta weight in regression analysis19, thus, a standardized factor loading would tell us how much an indicator would change, given a one unit change in the latent construct. In addition, each indicator includes a measuremen t error that explains the portion of the indicator that is measuring something else, othe r than what the latent variable is supposed to measure. Second there is the structural part of the model that represents the relationships between th e latent construct and the exogeno us causes. Such relationships are also explained by factor loadings, which in this case illustrate the change in the particular Capability given a unit change in a specific exogenous cause (e.g. race), holding the rest of variables constant. In the structural part of th e model, the exogenous variables are assumed to be free from measurement error, but each latent capability has an associated disturbance term, which is s upposed to measure the unexplained variance of the latent variable due to omitted causes (Bollen and Davis 2009; Schumaker and Lomax, 2004). Following Bollen and Davis (2009) notation the MIMIC model for both the Knowledge and Shelter Capabilities is formally represented as follows: 19. In regression analysis beta weights are the regression coefficients estimated from the standardized data, and are used to compare the relative importance of the different explanatory variables. A beta weight expresses the average change in the dependent variable when the independent variable changes one standa rd deviation from its mean, holding all other independent va riables constant or at their means (Garson 2008, 2009). 42
x (1) yy (2) Where (1) represents the structural part of the model that depi cts the relationships between each latent capability (Knowledge and Shelter) a nd the exogenous variables (e.g. race, age, marital status); is the matrix of factor loadings (beta weights) and is the vector of disturbances. This equation m easures the impact exogenous variables such as gender or race have on each capability dimension, while the disturbance term captures the portion of the capability that is not bei ng explained by those variables. Equation (2) is the measurement part of the model that depicts the relationships between the latent variable and each of its corresponding indicators These relationships are explained by the corresponding matrix of factors loadings .' syy that give the expected change in the indicators, given a unit change in the latent variable. Fo r instance, if the Knowledge Capability (the freedom an individual ha s to attain a given level of knowledge) is contracted given a unit change in one of the external causes, the in dicator variables (e.g. years of education) would be reduced as to refl ect such a reduction in the latent construct. Finally, gives the measurement error term for each indicator variable, which accounts for the portion of the indicators that are meas uring something else other than the latent construct. The measurement part of the model (2), takes into account the different indicator metrics. As there is a combination of con tinuous, categorical and dichotomous variables there is a need to specify non linear rela tionships when either a categorical or 43
dichotomous variable is presen ted. Hence, if an indicator is measured as a continuous variable, the equation (2) represen ts directly the relationship with the latent construct. On the other hand, following Muthn (1983, 45) and Krishnakumar and Ballon (2008, 996) the relationship in the measurement model (2) with dichotomous or categorical indicators is best expressed as follows: jy Represents a dichotomous variable in the indicators vector. For instance, living in an overcrowd household or not, where i identifies the individual and j the indicator. The correspondence between the indicator and the latent variable introduces a latent response variable ijyijy ~ such as that: iijjijijxbyy ~ (3) Where 1 0 ~ijyThis means that if in (3) in this case represents the latent capability dimension for shelter increases, and an individual loca ted in an overcrowded household could move to a non overcrowded household as a result of the increase in the capability dimension; such change is measured as going fromiy1 ~ ijy to 0 ~ ijy, in which case the factor loading j (beta weight) would pr esent a negative sign. In the case of an orde red categorical indicator with C categories, which will vary from 0, 1 to C -1; for instance, years of edu cation, which in our case goes from 0 to 21, we have the next expression: jy 44
iijjijijxbyy ~ (3) Where for cyijjcij cjy1 ~ and 1...2,1,0 C c j 0 and jc 1 This means that if in (3) is changed to represent the latent capability dimension for knowledge increases, an individual wi th say C=5 (5 years of education) could increase to C=6 or more as the result of the increase in the capability dimension. Such change is measured as going fromiy5 ~ ijy to6 ~ ijy, in which case the factor loading (beta weight) would present a positive sign depicting the positive relation between the Capability dimension and the corresponding indicator. Latent class involves the analysis of variance/covariance structures.20This means that although latent modeling can analyze means structures in specific models its primary aim is to assess the strength of the relationship between a given set variables and to explain as much as possible their va riability (Kline 2005). This analysis is 2211 2 y yrrR20. It is possible to have an intuitive meaning of analysis of covariance structures if we remember the relations behind the squared multiple corre lation coefficient, R2, in Multiple Regression analysis. The R2 tells how much variance in Y is explained, predicted and accounted for by the set of explanatory variables X1 and X2. For instance, for one dependent variable and two predictors, the R2 is: Where are the beta weights and their formal representation is as follows: 2 12 1221 11 r rrryy and 2 12 1212 21 r rrry y 1 yr2 yr12r, and and are the bivariate correlations am ong the explanatory and criterion variables. From here, it is possible to understand that regression analysis estimates a set of parameters (beta weights), from the correlations among the variab les, which in turn serve to construct the squared multiple correlation coefficient to explain as much variance as possible on Y given the correlation of the set of predictors X. Latent variable models devel op this same type of analysis but for latent and intervening variables (combination of path and factor analysis). 45
accomplished through the estimation of the model implied variance/covariance structure ( ) composed in our case of three covariance st ructures, and which is later compare to the sample covariance structure (S), to determ ine how well the model explains the variability patterns of the data. If the difference between these two st ructures, the sample and the model implied variance/covariance terms, is small enough; the model is say to be a reasonable depiction of the phenomenon analyz ed, and inference would be reliable. The first two variance/covariance terms we need to define are, one for the exogenous variables, denoted as and one for the disturbance terms, denoted as as follows: (5) This expression represents a matrix array of variance/covariance terms that measure the strength of the relationship between the set of exogenous variables (e.g. gender, race, marital status a nd region) plus the variance of the disturbance term, which as mentioned above is the portion of variance of the latent construct that is not explained by the set of exogenous variables. Second, from the measurement model we have the variance/covariance term of the measurement errors, denoted as These are the terms that designate the portion of the indicator variables that are measuring some thing else other than the associated latent capability and this matrix represent the covariance among them. Finally, with the above two expressions it is possible to construct the overall covariance structure of the specified model ( ): 46
')'(y y (6) This structure is the general expressi on of the variance/covariance term for the MIMIC model of latent Capabilities. After this structure has been es timated, it is compare to the variance/covariance struct ure of the sample, and if th e difference is close to zero the model presents a good fit, and allows the make reasonable inference from the phenomenon at hand. In addition, the following stochastic assu mptions must be met in order to estimate the model: 0 Expected value of disturbance must be zero 0 Expected value of measurement error must be zero Var, Variance of must be Var, Variance of must be Concretely, the estimation of a MIMIC model implies that the previous set of variance/covariance structures allow for the es timation of a set of pa rameters that denote the relationship between a given set of variab les, similar to beta weights in regression analysis. Once these parameters are estimated from the specified model, a sample variance/covariance structure is composed and is compared to the population matrix to 47
assess the extent to which the model specified by the researcher fits the data and allows one to make reasonable inferences. Path Diagram for MIMIC Model of Latent Capabilities By academic consensus, latent variable models are expressed graphically using path diagrams, which is also a more pedagogical approach to the understanding of this type of theoretical model. Their general aim is to s how graphically the causal relationships between the variables in a given model. Path diagrams have several rules or customary representations: Firstly, latent variables are represented with circles or ellipses; secondly, exogenous cau ses and observed variables ar e represented as squares; thirdly, disturbances from the latent construc ts and errors from the observed variables are drawn as small circles; and finally, arrows depict causal relati onships. A one-headed arrow expresses a one way causal relations hip and a two-headed arrow expresses a reciprocal relationship between any pair of variables. 48
Figure 1 Path Diagram for a general MIMIC model This path diagram describes the relationships modeled in a MIMIC model of latent Capabilities. It is ar gued that the exogenous variab les are influencing the Latent Capability, and at the same time the latent Capability determines the welfare a person is able to achieve in the given dimensions, fo r instance, achieved leve l of education. In addition, the error term Ei gives the portion of the indicators that are measuring something else other than the corresponding latent construct, while the disturbance term gives the portion of the latent Capability that is not caused by the exogenous variables. 49
Chapter 5. Results Integrating the theoretical approach and the statistical method, one can now present the estimated results and their co rresponding analysis. Th e next chart shows estimated results for the factor loadings of the structural model for the knowledge capability. These factor loadings composed th e matrix B in equation (1), and estimate the relationship between the exogenous causes and the latent construct. In other words, they estimate the influence that personal and househ old characteristics have over the freedom a person enjoys in achie ving the level of knowledge he or she wants. In addition, it is important to note that the sample size for this analysis is reduced from 85.145 individuals to 77.369 individuals given that there is not information on education for children under five years old. Al so, as it was mentioned above, for practical reasons, just the signif icant parameters at the 0.01 and 0.05 critical level are presented. Table 6 MIMIC model of Knowledge Capabi lity Achievement: Factor Loadings of the Structural Model Exogenous Causes Capability Dimension Knowledge Standardized C. U. Coefficient S. Error Gender -0.0853 -0.083 0.00315 Ethnic -0.0368 -0.358 0.00276 Poor -0.2721 -0.2648 0.00305 Urban/Rural -0.2432 -0.2367 0.00272 Working 0.3232 0.3145 0.0303 Famishing -0.0327 -0.0318 0.00282 50
The table presents the standardized coe fficient and the unstandardized coefficient with its corresponding standard errors (standardized coefficients do not have standard errors). For comparison purposes the standard ized coefficients are commonly used. They are interpreted in a similar manner to regr ession coefficients. For instance, all other things being equal, a statistically significant negativ e coefficient on a dichotomous indicator implies a lower m ean value of the knowledge capability for the groups identified by the exogenous indicator (H arrington 2009). The re sults for the MIMIC model of the knowledge capability show that the variables with higher explanatory power are, being poor, area and working status with factor loadin gs of -0.2721, -0.2432 and 0.3232 respectively. Here, a factor loading of -0.2721 indicates that a one unit change, which means going from non poor to poor, produ ces -0.2721 standard units of change from its mean in the latent factor when cont rolling by the rest of explanatory variables. Then, being poor and being in rural areas reduce significantly the freedom a person enjoys when choosing the level of knowledge he or she wants to attain; while working status has the contrary effect. The factor loading for ethnic membershi p, although it is low, is statistically significant at the 0.05 alpha le vel, and implies that when holding all other variables constant, Afro-Colombian (and Indigenous ) people would have a lower freedom for achieving the knowledge they want in relation to the reference group: non Afrodescendants. This can be e xpressed by saying that being black or indigenous is an individual characteristic that by itsel f reduces the set of choices people have. Furthermore, if the negative impact of be ing located in rural areas is summed up to 51
indicate negative influence of ethnic bac kground, it is possible to observe that the vulnerability of peoples of Af rican descent is endemic. In addition, it is important to highlight the negative im pact that gender has on the knowledge capability. If one r ecalls from table 4 that gender is measured as a dichotomous variable with 1 being male and 0 being female; the negative coefficient implies that males have lower capability achievement than females. This could be explained by the fact that in Colombia school enrollment rates for girls are higher than that of boys, which could be due to the na ture and magnitude of child labor in the country. Boys are usually sent into the labor market at an earlier stage of their life than girls. However, gender bias is still strong, but is more easily detected in the intrahousehold distribution of chores, and labor markets studies where income gaps persist between males and females. Subsequently, we present the results obt ained for the measurement part of the model (2) that depicts the relation between the latent construct and its corresponding indicators. The chart below illustrates the relationship between the latent knowledge capability and its functioning indicators asse ssed by the corresponding factor loadings. Again standardized coefficients ar e analyzed for comparison purposes. Table 7 MIMIC Model of Knowledge Capabili ty Achievement. Factor Loadings for the Measurement Model Knowledge Capability Standardized C. Coefficient S. Error Years of Education 0.9658 1.0 Illiteracy Rates -0.4109 -0.4222 0.0059 Technical Edu. For Work 0.3685 0.3787 0.00529 R2 0.3184 52
In this case what these loadings tell us is the standard unit ch ange of the indicator variable from its mean for each unit change of the latent construct. For instance, when there is an increase in the knowledge cap ability by finding a job or moving to urban areasas the results explained before show the achieved years of education would increases in 0.95 standard units from its mean. However, the contrary holds true also; if being a member of an ethnic minority decreases the freedom a person enjoys in attaining certain levels of knowledge, ach ieved years of education woul d be consistently lower in this segment of the population as it has been the case historically for Afro-Colombians. Given the observed indicators in the case of th e Illiterate rates the sign of the coefficient is consistent with the metric of the variable as this measures illiteracy as 1 and literacy as 0. Thus, an increase in the knowledge capability would de crease illiter acy by 0.4109 standard deviations from its mean, all others being constant. In this case, these results show that the freedom Afro-descendants en joy in choosing their knowledge level has been historically lower than th at of mestizos and whites. No t just for the fact of being poor as has been traditionally argued, but also for the fact of being black. They have been unable to attain higher levels of education as the result of cumulative effects of poverty and racial discrimination. When fitting a statistical model to a data set it is important to analyze how appropriate the fit statistics are, which tells the degree to which the model specified by the researcher is representing the given da ta properly (Schumack er and Lomax 2004). For this model fit indices appear well behaved, with the exception of the 2 index, which is too large. However, many authors have point ed out that a combination of reasonable 53
54indices by other measures allow one to pass over this chi squared as it is very sensitive to large sample sizes, which is the case with this example. The rest of indices perform reasonably well, which means that they are in the thresholds defined by the literature on this type of model. For instance the Root Mean Square Error of Approximation (RMSEA) is 0.0443 (RMSEA<0.06); Confiden ce limits for RMSEA are 0.0410-0.0443. The Comparative Fit Index (CFI) is 0.9819 (C FI>0.95); McDonalds (1989) measure of centrality is a fit index t-scaled to call between 0 and 1 (the cl oser to one the better fit the model has), and in our case is 0.9882. The Normed Fit Index (NFI) is 0.9818 (NFI>0.95). In sum, the usefulness of fit st atistics illustrates that the MIMIC model for the knowledge capability is appr opriate for the data, and allo ws one to assure that the inference made regarding the impact, that the set of exogenous variables on the knowledge capability has, is reliable enough; including the analysis made on the basis of the negative impact that ethnicity member ship has over the given capability. The next path diagram (see literal C, chapter IV) illustrates the relationships identified in the MIMIC model for the knowledge capability, which were explained by the previous analysis of factor loadings.
Figure 2 Path Diagram fo r the Knowledge Capability 55
As for the results of the structural pa rt of the MIMIC model for the Housing Capability, the next table s hows the factor loadings of the structural model or the relationships between the latent construct and its causal factors. Similar to the knowledge capability, standardized coefficients (w hich have no standard errors) and non standardized coefficients with their co rresponding standard errors are shown. Table 8 MIMIC model of Housing Capabili ty Achievement Factor Loadings of the Structural Model Exogenous Causes Capability Dimension Housing Standardized C. Coefficient S. Error Ethnic -0.1696 -0.1561 0.00307 Poor -0.0398 -0.0367 0.00441 Deciles 0.259 0.2716 0.00512 Area -0.5658 -0.5209 0.00478 Risk -0.127 -0.1169 0.0029 Similarly to the interpretati on of the factor lo adings of the exoge nous variables on the knowledge capability, all the coefficients are significant at 0.01 alpha level, but the coefficient for the poor variable becomes significant at the 0.05 alpha level. What is possible to observe here, is that all coefficients have signi ficant impact over the latent capability, and the variable which has the higher explanatory power is the variable area, which is measured as 0 for urban and 1 for rural areas (see table 4). Area has a standardized factor loading of -0.5658 which means that, when holding all other variables constant, going from urban to rural change s the housing capability negatively -0.5658 standard units from its mean. This is not surprising since our i ndicators are mainly indicators that refer to infrastructure; not only private infrastructure but also public ones, 56
and rural areas have a signifi cant backwardness in relation to urban areas in terms of infrastructure. The second explanatory variab le with high factor loading is income deciles. This is also not surprising since this capability is highl y correlated with the availability of economic resourcesthe highe r the decile the better the infrastructure available. The coefficient for the variable risk, which expresses the location of a household in a geographical area vulnerable to natural disaster s, presents a logical result. Thus, being located in a risk zone decrea ses the housing capability by 0.127 standard units from its mean, when holding all the other variab les constant. However, what is more important is the size, sign and significance of the coefficient for ethnic background, which implies that when holding all other variables constant, going from non Afro-descendant to Afro-descendant decreases the housing capability in -0.1696 standard units from its mean. These results suggest that after weighting the impact that being poor (havi ng lower income) and being located in rural areas have over the freedom a person enjoyed in achieving his or her valued level of knowledge and shelter, there is still a persis tent negative impact associated with the ethnic background of the indivi dual. Afro-descendants have consistently lower freedom in choosing the levels of welfare they want and have reason to value in both the housing and knowledge dimensions; not only for their condition of poverty and their mainly rural location, but also because they are black. Similarly to the results presented for the knowledge capability, table eight shows the results for the measurement model pa rt of MIMIC for shelter capability. 57
58Table 9 MIMIC Model for Shelter Capabili ty Achievement. Factor Loadings for the Measurement Model Housing Capability Standardized C. Coefficient S. Error Basic Services 0.8425 0.9150 0.00693 Dwelling 0.7188 0.7807 0.7807 Overcrowd -0.4604 -0.5 R2 0.7249 The factor loadings or standardized coe fficients for the measurement part of the MIMIC model for the housing capability show that an increase of one standard unit in the housing capability would increase the availability of basic services by 0.8425 standard units from its mean. Thus, if an individual m oves from rural to urban areas or from risk zones to zones free from risk of natural disa sters, his or her capab ility or freedom to attain higher levels of welfare in the hous ing dimension increases. Whereas the loading for overcrowding depicts the oppos ite relation, that an increase in the housing capability by one standard unit would have a decrease in the ove rcrowding variable of -0.4604 standard deviations from its mean. Hence, having more space or less people living in the same household would determine higher levels of achieved welfare in the shelter dimension. Finally, the next di agram depicts the relationships estimated by the model.
Figure 3 Path Diagram for the Shelter Capability 59
In addition, as the R2 for this model is higher showing a highly explanatory power, one has the result that the fit indexes reflect such power. The RMSEA is equal to 0.01008<0.06; CFI equal to 0.9921> 0.95; NFI equal to 0.9920. Mc Donalds (1989) is equal to 0.9504 and NFI equal to 0.9920. To summarize, it is possible to observe that after controlling by relevant individual, household and exte rnal factors such location, the coefficient for ethnic membership is significant and negative. This implies that at any average level of the knowledge and shelter capability dimensions, Afro-descendants in Colombia enjoy less freedom in achieving a wanted quality of life. Also, it is important to highlight that, after having included outcome variables from a so cial mobility perspective such as working status and being poor as exoge nous causes of capabilities, ethnic background continues to be significant. This suggests that in fact there is a component in welfare analys is of racial gaps in Colombian that cannot be assessed from class disparity approaches and their specific concentration on income-based measures. This is paramount to saying that from a Capability Approach perspe ctive, individual freedoms fo r Afro-Colombians are being constrained by race-based disc rimination. Consequently, it is their freedom and agency which must be enhanced and not merely thr ough the access to resources that they have. This is not to deny the importance and instrumentality that the command over resources has in supporting the achievement of welfare. What is really importa nt is to support the argument that there are other aspects of welfare that ca nnot be accounted for solely by income and resource access measures (R obeyns 2003). Finally, although one cannot identify the specific causal mechanisms by which race-based discrimination is taking 60
61place, these results, also support the argument that more reliable and systematical collection of data of ethnic and racial minorities across dimensions is needed.
Chapter 6. Conclusions For Sen (1999), poverty can be a trap in which those who do not have access to education, health and food, are unable to overcome their pre carious circumstances. Also, Duncan (1969) argues that pov erty is not a phenomenon that affects different social groups alike; on the contrary, historically marginalized so cial groups are affected in greater proportions by lack of access to resources such as education, health and employment opportunities. Specifi cally, this author argues th at: ...Negroes are poor mainly because they are Negroes and are de fined and treated as such by our society and their poverty stems largel y not from the legacy of poverty bu t from the legacy of race (p. 87). Consequently, black people have less po ssibility to prevail over conditions of poverty and exclusion because social practices of racial discrimination persist and are rooted in every day social in teraction. Furthermore, racial discrimination prevents the achievement of welfare not only to those belo w poverty lines, but also to those who are not under considered poor by income poverty measures. From this perspective, the descriptive an alysis of this study in conjunction with the results of the empirical a pplication of the Capability A pproach, suggest that racial discrimination plays an important role in limiting Afro-descendants capabilities or individual freedom. Although, cumulative effects of class disparities play a major role in determining their lower outcomes as the load ings for the poor and the income decile variable shows; cumulative effects of racial discrimination are taking place as well, as 62
suggests by the factor loading of the ethnic variable. Furthermore, the analysis of group disparities based merely on social stratificati on theories that uses income-base measures as their main explanatory variable is hinderi ng the political debate regarding the impact that racial discrimination have on individual welfare. Furtherm ore, this analysis is also misguiding the design and implementation of pub lic policy toward the mere increases of resource access. Wider considerations regarding human flourishi ng are not considered. For instance, political participation on an equal basis, the impor tance of commanding social respect on the basis of ones citizenship and the possib ility to claim justice when ones rights are violated are important missing aspects of the evaluation of Afrodescendants welfare. These missing aspects cannot be properly account for by standard analysis of welfare measured as income levels. As a consequence, the multidimensional natu re of the Capability Approach calls for the recognition and analysis of racial di scrimination as a separate dimension when assessing human welfare of indi viduals member of ethnic and racial minorities, in this case of Afro-Colombians. This inclusion would result in more accurate design of public policy toward the expansion of freedoms and not just of resource access. From this perspective, the social and political denial of the very ex istence of the phenomenon is limiting the collection of relevant data and the design of relevant public policy directed to address it. Thus, a good starting point to ward the expansion of Afro-Descendants freedom in Colombia would be end of the stat e of denial in which the society has lived regarding the existence of racial discrimination and the major role that it has played in the marginalization of this group. In addition and on urgent basis, more emphasis must be placed on the judicial and legal enforcement mechanisms to prevent and punish racial 63
discrimination practices. Without them, the guarantee of basic and constitutional rights for Afro-Colombians is an unreachable goal, and they would continue living as an underclass in the midst of modernity. 64
References Alkire, S. 2005. Valuing freedoms: Sen's Capabil ity Approach and Poverty Reduction Oxford University Press, USA. Anand, S., and A. K Sen. 1994. Human Development Index: Methodology and Measurement United Nations Development Programme (UNDP). Andrews, George Reid. 2004. Afro-Latin America, 1800-2000 Oxford University Press US. Arndt, C., H. Strotmann, and J. Volkert. D ynamic Aspects of Capability Deprivation on the micro-level an Empirical Essessment for Germany using the SMOP Method. Arocha, J. 1998. Inclusion of Afro-Col ombians: Unreachable National Goal? Latin American Perspectives : 70. Ballet, J., J. L Dubois, and F. Mahieu. 2007. Responsibility for Each Other's Freedom: Agency as the Source of Collective Capability. Journal of Human DevelopmentBasingstoke8, no. 2: 185. Barbary, O., H. F Ramirez, and F. Urrea. 2002. Identidad y ciudadana Afro-Colombiana en la regin Pacfica y Cali: Elementos esta dsticos y sociolgicos para el debate de la cuestin negra en Colombia. Estudos Afro-Asiticos Bello, and M. Rangel. 2002. La equidad y la exclusin de los pueblos indgenas y Afro-Descendientes en Amr ica Latina y el Caribe. Revista de la CEPAL 76: 39 54. Bharadwaj, A. S, V. Sambamurthy, and R. W Zmud. 1999. IT capabil ities: theoretical perspectives and empirical operationalization. In Proceedings of the 20th international conference on Information Systems 378. Grewal, Rajdeep, Joseph A. Cote, and Hans Baumgartner. 2004. Multicollinerarity and Measurment Error in Sitructural Equa tion Model: Implicantions for Theory Testing. Journal of Marketing Science 23(4), 519-529. Bhattacharya, J. Analyzing Women Well-bei ng Data with a MIMIC Model: A Districtlevel study in West Bengal, India. 65
Bollen, K. A, and W. R Davis. 2009. Causal Indicator Models: Identification, Estimation, and Testing. Structural Equation Modeling: A Multidisciplinary Journal 16, no. 3: 498. Clark, D. A. 2005. The Capability Approach : Its Development, Critiques and Recent Advances. Global Poverty Research Group : 3. Consultoria para los Derechos y el De splazamiento (CODHES). 2008. Afrocolombianos Desplazados. Un Drama sin Tregua. http://www.codhes.org/index.php?option=com_content&task=view&id=157 Cohen, G. A. 1993. Amartya Sen's Unequal World. Economic and Political Weekly 28, no. 40 (October 2): 2156-2160. Comim, F. 2001. Operationalizing Amartya Sen' s Capability Approach. Paper presented at the Conference Justice and Poverty: Examining Amartya Sen's Capability Approach, 5-7 June, Cambridge, United Kingdom. CONPES. 2004. 3310 de 2004 Poltica de Accin Afirmativa para la Poblacin Negra o Afro-Colombiana DNP. Cossio, F. V, P. Murillo, P. M.M Zapata, M. E.B Rendn, E. R Ariza, S. General, M. M Bohrquez, et al. Documento Propuesta para la Formulacin del Plan Integral de Largo Plazo Poblacin Negra/Afro-Col ombiana, Palenquera y Raizal. DEFINICIONES, O. Y. 1991. La Ley 70 de las Comunidades Negras. Amrica Negra: 197. Deneulin, Sverine. 2006. The Capability Approach and the Praxis of Development Basingstoke [England]: Palgrave Macmillan. Deneulin, Sverine, Mathias Ne bel, and Nicholas Sagovsky. 2006. Transforming unjust structures. Springer, July 19. Di Tommaso, M. L. 2007. Children Capabilities: A Structural Equation Model for India. Journal of Socio-Economics 36, no. 3: 436. Duncan, D. O. 1696. Inheritance of Poverty or Inheritance of Race ? in Moynihan, D. P. and C. S. Schelling (eds.). 1969. On understanding poverty; pe rspectives from the social sciences. New York, Basic Books. Escobar, A. 2003. Displacement, Development, and Modernity in the Colombian Pacific. International Socia l Science Journal 55, no. 175: 5. Evans, P. 2002. Collective Capabilities, Cu lture, and Amartya Sen's Development as 66
Freedom. Studies in Comparative International Development (SCID) 37, no. 2: 54. Garson, D. 2008. Path Analysis. From Statnotes: Topics in multivariate Analysis. Retrieved March 25, 2009. http://faculty.chass.ncsu.edu/garson/PA765/path.htm Gasper, D. 2004. The Ethics of Development Edinburgh University Press Edinburgh. Gerring, John. 2001. Social Science Methodology Cambridge University Press. . 2007. Case Study Research Cambridge University Press. Grace Miller, M. 2004. The Rise and Fall of the Cosmic Race. The Cult of Mestizaje in Latin America. University of Texas Press. Harrington, D. 2008. Confirmatory Factor Analysis Oxford University Press. Hicks, D. A. 2002. Gender, Discrimination, a nd Capability: Insights from Amartya Sen. The Journal of Religious Ethics : 137. Hogan, D. P, and D. L Featherman. 1977. Racial stratification and so cioeconomic change in the American North and South. American Journal of Sociology 83, no. 1: 100. Jreskog, K. G, E. B Andersen, P. Laake, D. R. Cox, and T. Schweder. 1981. Analysis of Covariance Structures [with Discussion and Reply]. Scandinavian Journal of Statistics : 65. Jreskog, K. G, and A. S Goldberger. 1975. Estimation of a Model with Multiple Indicators and Multiple Causes of a Single Latent Variable. Journal of the American Statistical Association : 631. Kakwani, N., and J. Silber. 2008. Introduc tion: Multidimensional Poverty Analysis: Conceptual Issues, Empirical Illu strations and Policy Implications. World Development 36, no. 6: 987. Kline, Rex B. 1998. Principles and Practice of St ructural Equation Modeling Methodology in the social scien ces. New York: Guilford Press. Krishnakumar, J. 2007. Going Beyond Functioni ngs to Capabilities: An Econometric Model to Explain and Estimate Capabilities. Journal of Human Development 8, no. 1: 39. Krishnakumar, J., and A. L Nagar. 2008. On Exact Statistical Properties of Multidimensional Indices based on Princi pal Components, Factor Analysis, MIMIC and Structural Equation Models. Social Indicators Research 86, no. 3: 481. 67
Krishnakumar, Jaya, and Paola Ballon. 2008. Esti mating Basic Capabilities: A Structural Equation Model Applied to Bolivia. World Development 36, no. 6 (June): 9921010. Kuklys, W. 2005. Amartya Sen's Capability Approach: Theoretical Insights and Empirical Applications Springer Verlag. Lasso, Marixa. 2007. Myths of harmony. Race and Republicanism during the Age of Revolution, Colombia,1795-1831 University of Pittsburgh Press. Majumdar, M., and S. Subramanian. 2001. Capability Failure and Group Disparities: Some Evidence from India for the 1980s. Journal of Development Studies 37, no. 5: 104. Marasinghe, M. G, and W. J Kennedy. 2008. SAS for Data Analysis: Intermediate Statistical Methods Springer Verlag. Martinetti, E. C. 2000. A multidimensional a ssessment of well-being based on Sens functioning approach. Rivista Internazionale di Scienze Sociali 2: 207. Medina, C. A. 2002. Oferta la boral en Colombia de acuerdo al color de la piel. Cede, Facultad de Economa, Universidad de los Andes Moynihan, D. P. and C. S. Schelling (eds.). 1969. On understanding poverty; perspectives from th e social sciences. New York, Basic Books. MULTICULTURAL, H. U.N.A.N.P.Y. DEPARTAMENTO NACIONAL DE PLANEACION. Muthen, B. 1983. Latent Variable Structural Equation Modeling with Categorical Data. Journal of Econometrics 22, no. 1. National Research Council of the National Academies. 2004. Measuring Racial Discrimination Washington, D.C: National Academies Press. Nef, Jorge, and International Develo pment Research Centre (Canada). 1999. Human Security and Mutual Vulnerability IDRC. Nelson, E. 2008. From Primary Goods to Capa bilities: Distributive Justice and the Problem of Neutrality. Political Theory 36, no. 1: 93. Nussbaum, Martha C. 2000. Women and Human Development: the Capabilities Approach Cambridge: Cambridge University Press. Nussbaum, M. C. 2003 .Capabilities as F undamental Entitlement: Sen and Social Justice. Feminist Economics, 9(2-3), 33-59. 68
Nussbaum, M. C, and A. K Sen. 1993. The quality of life Oxford University Press, USA. Oaxaca, R. L, and M. R Ransom. 1999. Iden tification in Detailed Wage Decompositions. Review of Economics and Statistics 81, no. 1: 154. Observatory on Racial Discrimination. 2008. Racial Discrimination and Human Rights in Colombia. A Report on the Situation of the Rights of Afro-Colombians. Ediciones Uniandes, Bogot. Portilla, D. A. 2003. Mercado Laboral y Disc riminacion Racial. Una Aproximacin para Cali. Junio Documento CEDE, Universidad de los Andes. Quinonez, S. A. 2007. Los Afrocolombianos: Entr e la Retorica del Multiculturalismo y el Fuego Cruzado del Destierro. The Journal of Latin American and Caribbean Anthropology 12, no. 1: 213. Rawls, J. 1971. A Theory of Justice. Cambridge, Mass., Belknap Press of Harvard University Press Reiter, Bernd and Gladys Mitchell (Eds). 2009. Brazil's New Racial Politics. Lynne Rienner Publisher. (Grewal, Cote and Baumgartner 2004) Rivas, N. Y, T. H Saa, and C. E Agudelo. 2000. Impactos de la Ley 70 y Dinmicas Polticas Locales de las Poblaciones Afrocolombianas: Estudios de Caso. Documento de Trabajo 50.Universidad del Valle: Facult ad de Ciencias Sociales y Econmicas Rivera, F. 2001. Migrantes y racismo en Amrica Latina: dimensiones ocultas de realidades complejas. Versin final de investigacin encomendada por el Instituto Interamericano de Derechos Humanos y el BID. Quito. Robeyns, I. 2003. Sen's Capability Approach and Gender Inequality: Selecting Relevant Capabilities. Feminist Economics 9, no. 2: 61. Roche, J. M. 2008. Monitoring Inequali ty among Social Groups: A Methodology Combining Fuzzy Set Theory and Principal Component Analysis. Journal of Human Development 9, no. 3: 427. Saith, R. 2001.Capabilities: the Co ncept and its Operationalisation. QEH Working Paper Series 66 Queen Elizabeth House, University of Oxford. Schiller, B. R. 1970. Stratifie d Opportunities: The Essence of the" Vicious Circle". American Journal of Sociology : 426. . 1971. Class discrimination vs racial discrimination. The Review of Economics 69
and Statistics : 263. Schumacker, Randall E. 2004. A Beginner's Guide to Stru ctural Equation Modeling 2nd ed. Mahwah, N.J: Lawrence Erlbaum Associates. Seth, S. 2009. Collecting capability data on th e basis of existing surveys: The case of India. (Publish in Maitreyee the tri-annual ebulletin of the Human Development and Capability Association, June) Sen, A. 1976. Poverty: an ordinal approach to measurement. Econometrica: Journal of the Econometric Society : 219. . 1979. Equality of what? The Tanner Lecture on Human Values. Stanford University, May 22: 1979. ______.1990.Development as Capability Expansi on., in Keith Griffin and John Knight (eds), Human Development and the Inte rnational Development Strategy for the 1990s, London: Macmillan, pp. 41.58. . 1985a. Well-being, agency and fr eedom: the Dewey lectures 1984. The Journal of Philosophy : 169. Sen, A. K. 1985b. Commodities and capabilities. North-Holland Amsterdam. Sen, A. K, and J. E Foster. 1973. On economic inequality Oxford University Press, USA. Sen, A. K. 1999. Development as Freedom 1st ed. New York: Knopf. Skidmore, Thomas E. 1993. Black into White: Race and Nationality in Brazilian Though. Duke University Press, reprinted. Telles, E. E. Incorporating Race and Ethnicity into the UN Millennium Development Goals. Urrea, F. 2006. La Poblacin Af rodescendiente en Colombia. Los Pueblos Indgenas y Afrodescendientes en Amrica Latina y el Caribe. Santiago de Chile: CEPAL. Urrea, F., C. Vifara. 2006. Efectos de la R aza y el Gnero en Logro Educativo y Status Socio-Ocupacional para Tres Ciudades de Colombianas. Revista Desarrollo y Sociedad, segundo semestre 2006:115-163. Urrea, F., C. Vifara, H. F Ramrez, and E. Gmez, S. Vlez y F. Ruiz. 2005. Pobreza y Grupos tnicos en Colombia: Anlisis de sus Factores Determinantes y Lineamientos de Poltica para su Reduccin Informe final presentado al Departamento Nacional de Planeacin, DNP 70
71Wade, P. 1995. The Cultural Poli tics of Blackness in Colombia. American Ethnologist : 341. . 2002. Introduction: the Colombia n Pacific in perspective. Journal of Latin American Anthropology 7, no. 2: 2. Wainryb, Cecilia, Judith G. Smetana, and Elliot Turiel. 2007. Social development, social inequalities, and social justice CRC Press, October 1.
Appendix A Figure 4 Distribution of Afro-Colombian Population by their Location on the National Territory21 21. Observatory for Human Rights in Colo mbia. Information Source: Official Map of the Territorial Agencies IGAC National System of Information for Lands of Black Communities Technical Commission Law 70 of 1993. National Plan for the Development of th e Afro-Colombian P opulation, DNP, 8. 72